Skip to content

splicelogic

Translate differential-transcript-usage (DTU) results into interpretable exon-level splice events — a Bioconductor package from the Love lab that takes DTU output (transcript-level effect sizes between conditions) and recasts the per-transcript signal as concrete pairwise splice events (cassette exon, alt-5'/3' SS, intron retention, etc.) so an analyst can see what splicing change drives the DTU call rather than just which transcripts shifted.

  • Maintainer: Beatriz Campillo Miñano (Love lab, UNC-Chapel Hill) — beatrizcampillo29@gmail.com
  • Bioconductor: splicelogic (submitted Mar 2026, released in subsequent Bioc cycle)
  • Source: github.com/thelovelab/splicelogic
  • Vignette: splicelogic vignette
  • License: MIT + file LICENSE
  • Status at GBCC2025: preview talk (the package was submitted to Bioconductor in March 2026; the June 2025 talk was a prototype demo of the DTU-to-splice-event interpretation)

Talk at GBCC2025

  • Speakers: Beatriz Campillo Miñano, Michael Love (UNC-Chapel Hill)
  • Day / session: Day 2 (Tue Jun 24, 2025) — Oral Session 2B, Grace Auditorium (chair: Vincent Carey)
  • Talk title: "Visualizing and analyzing differential transcript usage"
  • Slides: TBD
  • Video: GBCC2025 playlist
  • Abstract / schedule: GBCC2025 program

What it does

splicelogic takes the output of a DTU analysis (e.g. from DRIMSeq, fishpond, or another transcript-level test) and resolves the transcript-level effect into pairwise exon-level comparisons. For each pair of transcripts that shift in relative abundance, it classifies the structural difference as a discrete splicing event class (cassette exon, alt-5' SS, alt-3' SS, intron retention, mutually exclusive exons, alt-first/last exon) and produces visualizations that make the underlying splice change inspectable. The aim is to bridge the gap between "transcript X went up while transcript Y went down" and "exon E is being included more often in condition A".

How it works

Core idea. Rather than operate on splice junctions, splicelogic works on whole transcript structures — each transcript and all its exons carry the DTU effect estimate forward, and pairwise up-vs-down transcript comparisons are decomposed into the discrete splicing event that distinguishes them.

Inputs / outputs. Input is a TxDb object (transcript annotation) plus a DTU result table with a per-transcript effect-size column. Output is a per-event table classifying each detected change as skipped exon, included exon, mutually exclusive exons, retained intron, or alternative splice site, stored as GRanges objects.

Key innovation. Method-agnostic upstream coupling: splicelogic accepts transcript-level effect estimates from any DTU tester — DRIMSeq, DEXSeq, satuRn, edgeR-diffSplice — so the interpretation layer is decoupled from the statistical-testing layer. Preserving full isoform context (rather than junction-level evidence) keeps the event call grounded in the annotated transcript model.

Parameters / API surface worth knowing. - prepare_exons(txdb, dtu_table, coef_col) — joins exon annotation to DTU results; coef_col names the effect-size column. - preprocess() — prepares exons for event detection. - find_se() — detects skipped exons specifically. - find_all_events() — runs the full event-classification sweep.

Canonical example. From the README workflow: load a TxDb, run DRIMSeq (or another DTU tester), pass its coefficient table through prepare_exons() → preprocess() → find_all_events() to get a per-event classification table tying each annotated transcript-pair shift to a specific splicing event class.

Where it fits in the corpus

  • AACR 2026: axis = bioinfo / AI methods; DTU interpretation matters for any cancer-splicing dossier (oncogene isoform switching, neoantigen-from-cryptic-splice work)
  • EuroBioC 2025: no direct counterpart on splicing interpretation
  • fishpond / DRIMSeq: the upstream DTU testers — splicelogic consumes their output rather than replacing them
  • ISMB 2026: scaffold; splicing methods overlap likely
  • Nextflow Summit 2026: not mentioned

Notes

The package was submitted to Bioconductor in March 2026, so the GBCC2025 talk was effectively a preview of work-in-progress. The release timeline matters for citation accuracy: if a user references the package from GBCC slides, point them at the github repo (thelovelab/splicelogic) rather than the not-yet-existing Bioconductor URL for the talk date. Campillo joined the Love lab as a BCB PhD candidate in June 2025; this was an early-PhD presentation.